Question regarding floating geometry

[Matth]

Howdy!

I got a small question, what would be the most efficient way to model things: Connect as much as possible, or to have a lot of un-connected meshes?

As an example, let's say I model some spikes on a plane. Should I extrude thoose from the mesh, and keep it all as one mass, or keep the plane 1 poly, and simply sink the spikes down into it?

Lately it's starting to confuse me. I always believed having as much as possible as 1 mass a lot better, but as I got into high poly modelling, I get too much problems if I have everyting connected.

Thanks in advance!
Matth

[megalmn2000]

Depends on what you want to do with.

Sometimes, you might want to save triangles by using floating geometry, that's fine, but if your floating is too big, it might not help the texture because the ao map will use a lot of space. One of the art test I did was a failure, I used too much big floating geometry and half of my texture is waste by black shapes.
(Art test) Watercooler - Polycount Forum



If you look at my Mass Effect M8 gun, you might see some floating geometry. They are small, but help to keep the triangle count low. If I connect them all to my model, there will be lot of useless triangle.
Minh Nhat Le - Environment Artist

And last thing, floating are useful for animations too.
Ex: you might want the trigger of the gun to be animated, not to stay at the same place. Just need to detach it and put the pivot point at the right place and that's it!

Hope it help.

[kanga]

Depending, parts can be detached but the same model. For a continuous surface keep it one but you arnt talking about continuity so break em up, specially for hi poly.

[Matth]

Thanks for the replies!

So, am I correct if I'd say: Keep it as 1 piece ( connected) as long as it doesn't cause side effects?
Side effects as in totally screwed up topology, or random hard lines from support edges.

Cheers,
Matth

[RMachucaA]

Theres also a few drawbacks to floating poly's. Depending on the 3d engine being used, it can cause issues with realtime shadows, z-buffer issues and lastly, if the mesh is to be deformed\animated you will definitively get some interpolations.

So if its a tiny little part that isnt that crucial to the integrity of the model, go ahead and make it float, otherwise its just good practice to make it part of the mesh.

[Rik V]

Instead of limiting yourself with a "rule" try to think in terms of practicality.

You could box model a table with legs but it would be pretty difficult and inefficient to to do the same if you model a vehicle like a tractor or a game level (especially when in enviroment modelling you use a lot of modular pieces). Why model each nut, door hinge assembly or window frame when you can just do one and copy array (assuming they're all identical).

I also think it pointless to worry about topology if you're modelling a high when your next step is to remesh or retopologize and use normal maps. Hi-res + good topology will matter if you model for movies however.

[Matth]

The thing is, high poly modelling a big mesh gets its drawbacks. It's not the first time I model something for a few houres, then notice the edge of the cylinder shows, since they did not get enough geometry to start with. Now I got 2 choices:

A) Remodel a few houres of work, killing my motivatio.
or
B) Make the cylinder shape loose, and take away any support edges that accidently went that way.

Hence I was wondering: is option B) a good option, or not?

Cheers,
Matth